Development And Application Of RH Refining Furnace Model Based On Machine Learning

With the development of science and technology,various vacuum pumps have come out one after another.At the same time,modern science and technology have also put forward higher requirements for the quality of materials.In order to reduce costs and improve product quality,various vacuum refinings have been produced one after another.RH has quickly become the mainstream of vacuum refining technology with its outstanding characteristics.RH vacuum processing technology has become a high-efficiency,large amount of molten steel and good refining effect The core of modern steel companies producing special steel types,especially ultra-pure steelBased on years of operating data and experience of a copper plant in Wuhan,this paper conducts steelmaking equipment construction and parameter guidance at the Bilai National Iron and Steel Plant in India,and adds a smart steel refining process control system model,mainly including temperature prediction models,alloys Feeding model and decarburization treatment model.The main research contents and conclusions of this article are as follows:Aiming at the temperature model,the factors that have a higher degree of temperature were screened first.Based on the factors,multiple linear regression method was used for modeling,processing and forecasting.The disadvantages of the method that did not combine quantitative analysis and qualitative analysis were analyzed.The case-based reasoning method,but this method averages the effect weights of the influencing factors,and then considers using particle swarm optimization weights to finally improve the accuracy and use it in actual production.For the alloy feeding model,a dynamic database of alloy element yields is first established,and then processed based on molten steel information,silo data,and actual production requirements using linear regression methods.Although it can meet the requirements of steel element composition,its operating environment is harsh There are also defects that the target controllable element only controls the element's target lower limit value,and then the non-linear programming method is used to further optimize the modeling method,so that the element is closer to the target value,and the cost must be controlled.For the decarburization treatment model,first analyze the decarbonization thermodynamics and various influencing factors,and then train the BP neural network model based on the existing historical data.Analyze the impact of the number of layers on the model accuracy to select the optimal structure.The rate is different in different time periods,so the genetic algorithm is used to divide the period,and finally the decarbonization curve is obtained.